Merge pull request #205 from QuanMPhm/187/acct_quota

Allow direct communication to Openshift Quota API

Author: knikolla

ci/run_functional_tests_openshift.sh

@@ -17,6 +17,8 @@ fi
 export DJANGO_SETTINGS_MODULE="local_settings"
 export FUNCTIONAL_TESTS="True"
 export OS_AUTH_URL="https://onboarding-onboarding.cluster.local"
+export OS_API_URL="https://onboarding-onboarding.cluster.local:6443"
+
 
 coverage run --source="." -m django test coldfront_plugin_cloud.tests.functional.openshift
 coverage report

src/coldfront_plugin_cloud/attributes.py

@@ -19,6 +19,7 @@ class CloudAllocationAttribute:
 
 
 RESOURCE_AUTH_URL = 'Identity Endpoint URL'
+RESOURCE_API_URL = 'OpenShift API Endpoint URL'
 RESOURCE_IDENTITY_NAME = 'OpenShift Identity Provider Name'
 RESOURCE_ROLE = 'Role for User in Project'
 
@@ -33,6 +34,7 @@ class CloudAllocationAttribute:
 
 RESOURCE_ATTRIBUTES = [
     CloudResourceAttribute(name=RESOURCE_AUTH_URL),
+    CloudResourceAttribute(name=RESOURCE_API_URL),
     CloudResourceAttribute(name=RESOURCE_IDENTITY_NAME),
     CloudResourceAttribute(name=RESOURCE_FEDERATION_PROTOCOL),
     CloudResourceAttribute(name=RESOURCE_IDP),

src/coldfront_plugin_cloud/management/commands/add_openshift_resource.py

@@ -17,6 +17,10 @@ def add_arguments(self, parser):
                             help='Name of OpenShift resource')
         parser.add_argument('--auth-url', type=str, required=True,
                             help='URL of the openshift-acct-mgt endpoint')
+        parser.add_argument('--api-url', type=str, required=True,
+                            help='API URL of the openshift cluster')
+        parser.add_argument('--idp', type=str, required=True,
+                            help='Name of Openshift identity provider')
         parser.add_argument('--role', type=str, default='edit',
                             help='Role for user when added to project (default: edit)')
 
@@ -37,6 +41,18 @@ def handle(self, *args, **options):
             resource=openshift,
             value=options['auth_url']
         )
+        ResourceAttribute.objects.get_or_create(
+            resource_attribute_type=ResourceAttributeType.objects.get(
+                name=attributes.RESOURCE_API_URL),
+            resource=openshift,
+            value=options['api_url']
+        )
+        ResourceAttribute.objects.get_or_create(
+            resource_attribute_type=ResourceAttributeType.objects.get(
+                name=attributes.RESOURCE_IDENTITY_NAME),
+            resource=openshift,
+            value=options['idp']
+        )
         ResourceAttribute.objects.get_or_create(
             resource_attribute_type=ResourceAttributeType.objects.get(
                 name=attributes.RESOURCE_ROLE),

src/coldfront_plugin_cloud/management/commands/validate_allocations.py

@@ -183,7 +183,7 @@ def handle(self, *args, **options):
                 )
                 continue
 
-            quota = allocator.get_quota(project_id)["Quota"]
+            quota = allocator.get_quota(project_id)
 
             failed_validation = Command.sync_users(project_id, allocation, allocator, options["apply"])
 

src/coldfront_plugin_cloud/openshift.py

@@ -36,12 +36,12 @@ def clean_openshift_metadata(obj):
     return obj
 
 QUOTA_KEY_MAPPING = {
-    attributes.QUOTA_LIMITS_CPU: lambda x: {":limits.cpu": f"{x * 1000}m"},
-    attributes.QUOTA_LIMITS_MEMORY: lambda x: {":limits.memory": f"{x}Mi"},
-    attributes.QUOTA_LIMITS_EPHEMERAL_STORAGE_GB: lambda x: {":limits.ephemeral-storage": f"{x}Gi"},
-    attributes.QUOTA_REQUESTS_STORAGE: lambda x: {":requests.storage": f"{x}Gi"},
-    attributes.QUOTA_REQUESTS_GPU: lambda x: {":requests.nvidia.com/gpu": f"{x}"},
-    attributes.QUOTA_PVC: lambda x: {":persistentvolumeclaims": f"{x}"},
+    attributes.QUOTA_LIMITS_CPU: lambda x: {"limits.cpu": f"{x * 1000}m"},
+    attributes.QUOTA_LIMITS_MEMORY: lambda x: {"limits.memory": f"{x}Mi"},
+    attributes.QUOTA_LIMITS_EPHEMERAL_STORAGE_GB: lambda x: {"limits.ephemeral-storage": f"{x}Gi"},
+    attributes.QUOTA_REQUESTS_STORAGE: lambda x: {"requests.storage": f"{x}Gi"},
+    attributes.QUOTA_REQUESTS_GPU: lambda x: {"requests.nvidia.com/gpu": f"{x}"},
+    attributes.QUOTA_PVC: lambda x: {"persistentvolumeclaims": f"{x}"},
 }
 
 
@@ -77,7 +77,7 @@ def __init__(self, resource, allocation):
     def k8_client(self):
         # Load Endpoint URL and Auth token for new k8 client
         openshift_token = os.getenv(f"OPENSHIFT_{self.safe_resource_name}_TOKEN")
-        openshift_url = self.resource.get_attribute(attributes.RESOURCE_AUTH_URL)
+        openshift_url = self.resource.get_attribute(attributes.RESOURCE_API_URL)
 
         k8_config = kubernetes.client.Configuration()
         k8_config.api_key["authorization"] = openshift_token
@@ -146,20 +146,41 @@ def create_project(self, suggested_project_name):
         project_name = project_id
         self._create_project(project_name, project_id)
         return self.Project(project_name, project_id)
+    
+    def delete_moc_quotas(self, project_id):
+        """deletes all resourcequotas from an openshift project"""
+        resourcequotas = self._openshift_get_resourcequotas(project_id)
+        for resourcequota in resourcequotas:
+            self._openshift_delete_resourcequota(project_id, resourcequota["metadata"]["name"])
+
+        logger.info(f"All quotas for {project_id} successfully deleted")
 
     def set_quota(self, project_id):
-        url = f"{self.auth_url}/projects/{project_id}/quota"
-        payload = dict()
+        """Sets the quota for a project, creating a minimal resourcequota
+        object in the project namespace with no extra scopes"""
+
+        quota_spec = {}
         for key, func in QUOTA_KEY_MAPPING.items():
             if (x := self.allocation.get_attribute(key)) is not None:
-                payload.update(func(x))
-        r = self.session.put(url, data=json.dumps({'Quota': payload}))
-        self.check_response(r)
+                quota_spec.update(func(x))
+
+        quota_def = {
+            "metadata": {"name": f"{project_id}-project"},
+            "spec": {"hard": quota_spec},
+        }
+
+        self.delete_moc_quotas(project_id)
+        self._openshift_create_resourcequota(project_id, quota_def)
+
+        logger.info(f"Quota for {project_id} successfully created")
 
     def get_quota(self, project_id):
-        url = f"{self.auth_url}/projects/{project_id}/quota"
-        r = self.session.get(url)
-        return self.check_response(r)
+        cloud_quotas = self._openshift_get_resourcequotas(project_id)
+        combined_quota = {}
+        for cloud_quota in cloud_quotas:
+            combined_quota.update(cloud_quota["spec"]["hard"])
+
+        return combined_quota
 
     def create_project_defaults(self, project_id):
         pass
@@ -181,7 +202,7 @@ def reactivate_project(self, project_id):
     def get_federated_user(self, username):
         if (
             self._openshift_user_exists(username)
-            and self._openshift_get_identity(username)
+            and self._openshift_identity_exists(username)
             and self._openshift_useridentitymapping_exists(username, username)
         ):
             return {'username': username}
@@ -264,25 +285,84 @@ def _openshift_get_identity(self, id_user):
     def _openshift_user_exists(self, user_name):
         try:
             self._openshift_get_user(user_name)
-        except kexc.NotFoundError:
-            return False
+        except kexc.NotFoundError as e:
+            # Ensures error raise because resource not found, 
+            # not because of other reasons, like incorrect url
+            e_info = json.loads(e.body)
+            if (
+                e_info["reason"] == "NotFound"
+                and e_info["details"]["name"] == user_name
+            ):
+                return False
+            raise e
         return True
 
     def _openshift_identity_exists(self, id_user):
         try:
             self._openshift_get_identity(id_user)
-        except kexc.NotFoundError:
-            return False
+        except kexc.NotFoundError as e:
+            e_info = json.loads(e.body)
+            if e_info.get("reason") == "NotFound":
+                return False
+            raise e
         return True
 
     def _openshift_useridentitymapping_exists(self, user_name, id_user):
         try:
             user = self._openshift_get_user(user_name)
-        except kexc.NotFoundError:
-            return False
+        except kexc.NotFoundError as e:
+            e_info = json.loads(e.body)
+            if e_info.get("reason") == "NotFound":
+                return False
+            raise e
 
         return any(
             identity == self.qualified_id_user(id_user)
             for identity in user.get("identities", [])
         )
 
+    def _openshift_get_project(self, project_name):
+        api = self.get_resource_api(API_PROJECT, "Project")
+        return clean_openshift_metadata(api.get(name=project_name).to_dict())
+
+    def _openshift_get_resourcequotas(self, project_id):
+        """Returns a list of resourcequota objects in namespace with name `project_id`"""
+        # Raise a NotFound error if the project doesn't exist
+        self._openshift_get_project(project_id)
+        api = self.get_resource_api(API_CORE, "ResourceQuota")
+        res = clean_openshift_metadata(api.get(namespace=project_id).to_dict())
+
+        return res["items"]
+    
+    def _wait_for_quota_to_settle(self, project_id, resource_quota):
+        """Wait for quota on resourcequotas to settle.
+
+        When creating a new resourcequota that sets a quota on resourcequota objects, we need to
+        wait for OpenShift to calculate the quota usage before we attempt to create any new
+        resourcequota objects.
+        """
+
+        if "resourcequotas" in resource_quota["spec"]["hard"]:
+            logger.info("waiting for resourcequota quota")
+
+            api = self.get_resource_api(API_CORE, "ResourceQuota")
+            while True:
+                resp = clean_openshift_metadata(
+                    api.get(
+                        namespace=project_id, name=resource_quota["metadata"]["name"]
+                    ).to_dict()
+                )
+                if "resourcequotas" in resp["status"].get("used", {}):
+                    break
+                time.sleep(0.1)
+    
+    def _openshift_create_resourcequota(self, project_id, quota_def):
+        api = self.get_resource_api(API_CORE, "ResourceQuota")
+        res = api.create(namespace=project_id, body=quota_def).to_dict()
+        self._wait_for_quota_to_settle(project_id, res)
+    
+    def _openshift_delete_resourcequota(self, project_id, resourcequota_name):
+        """In an openshift namespace {project_id) delete a specified resourcequota"""
+        api = self.get_resource_api(API_CORE, "ResourceQuota")
+        return api.delete(namespace=project_id, name=resourcequota_name).to_dict()
+    

src/coldfront_plugin_cloud/tests/base.py

@@ -81,13 +81,15 @@ def new_openstack_resource(name=None, auth_url=None) -> Resource:
         return Resource.objects.get(name=resource_name)
 
     @staticmethod
-    def new_openshift_resource(name=None, auth_url=None) -> Resource:
+    def new_openshift_resource(name=None, auth_url=None, api_url=None, idp=None) -> Resource:
         resource_name = name or uuid.uuid4().hex
 
         call_command(
             'add_openshift_resource',
             name=resource_name,
             auth_url=auth_url or 'https://onboarding-onboarding.cluster.local',
+            api_url=api_url or 'https://onboarding-onboarding.cluster.local:6443',
+            idp=idp or 'developer',
         )
         return Resource.objects.get(name=resource_name)
 

src/coldfront_plugin_cloud/tests/functional/openshift/test_allocation.py

@@ -15,7 +15,8 @@ def setUp(self) -> None:
         super().setUp()
         self.resource = self.new_openshift_resource(
             name='Microshift',
-            auth_url=os.getenv('OS_AUTH_URL')
+            auth_url=os.getenv('OS_AUTH_URL'),
+            api_url=os.getenv('OS_API_URL'),
         )
 
     def test_new_allocation(self):
@@ -36,7 +37,8 @@ def test_new_allocation(self):
         allocator._get_project(project_id)
 
         # Check user and roles
-        allocator.get_federated_user(user.username)
+        user_info = allocator.get_federated_user(user.username)
+        self.assertEqual(user_info, {'username': user.username})
 
         allocator._get_role(user.username, project_id)
 
@@ -73,7 +75,8 @@ def test_add_remove_user(self):
         tasks.add_user_to_allocation(allocation_user2.pk)
         allocator._get_role(user.username, project_id)
 
-        allocator.get_federated_user(user2.username)
+        user_info = allocator.get_federated_user(user.username)
+        self.assertEqual(user_info, {'username': user.username})
 
         allocator._get_role(user.username, project_id)
         allocator._get_role(user2.username, project_id)
@@ -123,17 +126,16 @@ def test_new_allocation_quota(self):
         self.assertEqual(allocation.get_attribute(attributes.QUOTA_REQUESTS_GPU), 2 * 0)
         self.assertEqual(allocation.get_attribute(attributes.QUOTA_PVC), 2 * 2)
 
-        quota = allocator.get_quota(project_id)['Quota']
-        quota = {k: v for k, v in quota.items() if v is not None}
+        quota = allocator.get_quota(project_id)
         # The return value will update to the most relevant unit, so
         # 2000m cores becomes 2 and 8192Mi becomes 8Gi
         self.assertEqual(quota, {
-            ":limits.cpu": "2",
-            ":limits.memory": "8Gi",
-            ":limits.ephemeral-storage": "10Gi",
-            ":requests.storage": "40Gi",
-            ":requests.nvidia.com/gpu": "0",
-            ":persistentvolumeclaims": "4",
+            "limits.cpu": "2",
+            "limits.memory": "8Gi",
+            "limits.ephemeral-storage": "10Gi",
+            "requests.storage": "40Gi",
+            "requests.nvidia.com/gpu": "0",
+            "persistentvolumeclaims": "4",
         })
 
         # change a bunch of attributes
@@ -154,16 +156,16 @@ def test_new_allocation_quota(self):
         # This call should update the openshift quota to match the current attributes
         call_command('validate_allocations', apply=True)
 
-        quota = allocator.get_quota(project_id)['Quota']
+        quota = allocator.get_quota(project_id)
         quota = {k: v for k, v in quota.items() if v is not None}
 
         self.assertEqual(quota, {
-            ":limits.cpu": "6",
-            ":limits.memory": "8Gi",
-            ":limits.ephemeral-storage": "50Gi",
-            ":requests.storage": "100Gi",
-            ":requests.nvidia.com/gpu": "1",
-            ":persistentvolumeclaims": "10",
+            "limits.cpu": "6",
+            "limits.memory": "8Gi",
+            "limits.ephemeral-storage": "50Gi",
+            "requests.storage": "100Gi",
+            "requests.nvidia.com/gpu": "1",
+            "persistentvolumeclaims": "10",
         })
 
     def test_reactivate_allocation(self):
@@ -180,19 +182,17 @@ def test_reactivate_allocation(self):
 
         self.assertEqual(allocation.get_attribute(attributes.QUOTA_LIMITS_CPU), 2)
 
-        quota = allocator.get_quota(project_id)['Quota']
+        quota = allocator.get_quota(project_id)
 
-        # https://github.com/CCI-MOC/openshift-acct-mgt
-        quota = {k: v for k, v in quota.items() if v is not None}
         # The return value will update to the most relevant unit, so
         # 2000m cores becomes 2 and 8192Mi becomes 8Gi
         self.assertEqual(quota, {
-            ":limits.cpu": "2",
-            ":limits.memory": "8Gi",
-            ":limits.ephemeral-storage": "10Gi",
-            ":requests.storage": "40Gi",
-            ":requests.nvidia.com/gpu": "0",
-            ":persistentvolumeclaims": "4",
+            "limits.cpu": "2",
+            "limits.memory": "8Gi",
+            "limits.ephemeral-storage": "10Gi",
+            "requests.storage": "40Gi",
+            "requests.nvidia.com/gpu": "0",
+            "persistentvolumeclaims": "4",
         })
 
         # Simulate an attribute change request and subsequent approval which
@@ -201,17 +201,16 @@ def test_reactivate_allocation(self):
         tasks.activate_allocation(allocation.pk)
         allocation.refresh_from_db()
 
-        quota = allocator.get_quota(project_id)['Quota']
-        quota = {k: v for k, v in quota.items() if v is not None}
+        quota = allocator.get_quota(project_id)
         # The return value will update to the most relevant unit, so
         # 3000m cores becomes 3 and 8192Mi becomes 8Gi
         self.assertEqual(quota, {
-            ":limits.cpu": "3",
-            ":limits.memory": "8Gi",
-            ":limits.ephemeral-storage": "10Gi",
-            ":requests.storage": "40Gi",
-            ":requests.nvidia.com/gpu": "0",
-            ":persistentvolumeclaims": "4",
+            "limits.cpu": "3",
+            "limits.memory": "8Gi",
+            "limits.ephemeral-storage": "10Gi",
+            "requests.storage": "40Gi",
+            "requests.nvidia.com/gpu": "0",
+            "persistentvolumeclaims": "4",
         })
 
         allocator._get_role(user.username, project_id)